A novel approach to assess the dynamics of extra-chromosomal circular ribosomal DNA in human cells

Several nutrient-signaling pathways that extend life span have been described in model organisms. Thus, parallel and redundant signaling pathways that are similar across species might be subject to experimental manipulation. Here, we develop a PCR-based technique for testing the hypothesis that mitotic accumulation of extra-chromosomal ribosomal DNA circles might also determine life span in human cells. Using resveratrol, a phytochemical that counters age-related signs, we find treatment-dependent subcellular accumulations of extra-chromosomal 5S ribosomal DNA in human cell lines. These data suggest an association between DNA circles and intrinsic aging and demonstrate the utility of a PCRbased technique for studying the accumulation of dysfunctional molecules that promote senescence.


Introduction
Aging appears to be plastic and can be manipulated by genetic and nutritional intervention. For example, activation of the silent information regulator T1 (SIRT1) pathway increases the life span of model organisms such as yeast and mice. SIRT1 is a NAD + -dependent deacetylase that directly links transcriptional regulation to intracellular metabolism [1]. Among the signaling cues that activate SIRT1 pathways, is the polyphenol molecule resveratrol. SIRT1 activation by resveratrol triggers a broad range of transcription factors and co-regulators that mediate key mechanisms in the cell cycle, cell growth and apoptotic and autophagic programs of cell death [2][3][4]. Other intracellular mechanisms activated by SIRT1-dependent *Correspondence: Leheste  pathways are those related to DNA stability and DNA replication. For example, using the budding yeast (Saccharomyces cerevisiae) as a model for elucidating signaling pathways that control life span, Sinclair and Guarente [5] showed, that increased activity of SIR2 (the yeast ortholog of mammalian SIRT1) suppresses the gradual accumulation of extra-chromosomal circular DNA (eccDNA). As its name implies, e cc D NA are circular molecules propagated extra-chromosomally from repetitive and non-repetitive genomic regions of various species including, humans [6][7][8][9].
As in yeast, human ribosomal DNA (rDNA) is also organized in tandem rDNA repeats with five 43 kb rDNA (rna45s5) clusters encoding a large 45S rRNA precursor that is post-transcriptionally processed into 28S, 18S and 5.8S rRNAs [10,11]. 5S rRNA is encoded by a separate 5S rDNA (rn5s1) cluster with 100-150 copies of a 2.2kb rDNA tandem repeat [12,13]. Investigations into the genomic architecture of human rDNA clusters reveal significant meiotic re-arrangement of about 11% per generation per gene cluster [14], with 5SrDNA molecules undergoing significant replicative steps [7]. In humans, SIRT1 expression is also associated with rDNA re-arrangements suggesting a conserved composition and function of anti-aging pathways. However, there is little experimental evidence for these association pathways, in particular whether activation of SIRT1 or reduction of eccrDNA could be considered for the prevention of specific diseases. In addition, there is no sensitive platform for functional screening of human ec-crDNA which allows for the precise and accurate analysis underlying the formation of eccrDNA molecules. To address this technical limitation, we developed and validated a PCR-based protocol in combination with a nuclear transport assay for the quantitative analysis of ec-crDNA molecules. To further support its practical use, we tested whether this technique could identify the occurrence of additional eccDNA species in human cell lines.

Cell culture, drug treatment and collection of human cells
The following adherent human epithelial cell lines were propagated under standard culture conditions (37 o C, 5% CO 2 ) and as recommended by the manufacturer (ATCC, Manassas, VA, USA): HEK-293 (embryonic kidney, CRL-1673); SH-SY5Y (neuroblastoma, CRL-2266) and; MCF7 (mammary adenocarcinoma, HTB-22). Resveratrol treatment (50 µM, final concentra-tion, 6 h/48 h), was initiated at a cellular confluence level of about 70%. For nuclear transport studies, the lectin wheat germ agglutinin (WGA) was used to block nuclear transport (0.1 mg/ml, final concentration, 12 h) and the lectin, concavalin A (ConA), as control (0.1 mg/ml, final concentration, 12 h). At the end of a given treatment period, cells were washed with PBS, scraped off their dishes and collected for further analysis or stored (-80°C).

Resveratrol-dependent changes in gene expression using quantitative PCR (QPCR)
Following resveratrol treatment (48h), HEK-293, MCF7 and SH-SY5Y cells were collected and total RNA was prepared with the RNeasy RNA-isolation system/Qia shredder according to the manufacturer's specifications (Qiagen, Valencia, CA, USA). Following the determination RNA concentrations and integrity, complementary cDNA was generated with the Superscript III First Strand Synthesis System for RT PCR (Invitrogen, Carlsbad, CA, USA). QPCR was conducted on a Mastercycler ep gradient S (Eppendorf AG, Hamburg, Germany) using Power SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA, USA) in a total sample volume of 20 µl. Gene-specific DNA primers were either used as previously reported [15] or designed using the integrated DNA Technologies Primer Quest tool (IDT, Coralville, IA, USA). Expression was accessed for sirt1, rn5s1 (5SrRNA) and rna45s5 (45S pre-ribosomal RNA) using the following primers:

PCR amplification eccDNA for Southern blot or DNA sequencing analysis
Mixed nuclear and cytoplasmic eccDNA was isolated and purified from HEK-293 cells as described above. To specifically linearize 5S rDNA circles (single BamHI restriction site), one half of each sample (10 µl) was treated with BamHI for 4 h, 37°C while the other half was sonicated 3 times with 10s pulses on wet ice (30% maximum power, 2 mm tip; Cell Disruptor, Heat-Systems Ultrasonics, Plainview, NY, USA) for linearization/fragmentation. Samples were purified and eluted in a final volume of 10 µl. To generate blunt ends, samples were treated with T4-DNA polymerase and dNTPs (Roche Applied Science, Indianapolis, IN, USA) for 20 min at 16°C and immediately purified. EcoRI adaptor oligonucleotides were ligated to the linearized and blunt-ended DNA pieces for 18h at 16 o C and excess adaptors subsequently removed by gel filtration using Sephacryl S-400 spin columns according to the manufacturer's instructions (both: Universal Riboclone cDNA Synthesis System, Promega, Madison, WI, USA). Derivatives of eccDNA were then amplified using a high fidelity PCR system (Expand 20 kb plus PCR System, Roche Applied Science, Mannheim, Germany) and DNA primer matching the EcoRI adaptor sequence (forward and reverse primer: AAT TCC GTT GCT GTC G; Promega, Madison, WI, USA). Reactions destined for DNA cloning and sequencing were exclusively recruited from the sonicated samples and subject to PCR elongation reactions at 4 min. Reactions used for Southern blot analysis were recruited from sonicated as well as BamHI restricted samples and subject to 15 min PCR elongation reactions.

Identification and screening of eccD-NA in HEK-293 cells
Linearized and amplified eccDNA samples were briefly separated on ethidium bromide-stained 1.5% agarose gels from which a fragment cluster ranging from approximately 100 bp -4 kb was removed using the QI-Aquick Gel Extraction Kit (Qiagen,Valencia, CA, USA). Purified DNA was ligated into the pGEM-T Easy vector system (Promega, Madison, WI, USA) and transformed into chemically competent E. coli DH5α (Invitrogen/ Life Technologies, Carlsbad, CA, USA). Plasmid DNA from ampicillin-resistant and beta-galactosidase negative trans-formants (blue-white screening) was treated with EcoRI restriction endonuclease for 1 h at 37°C and analyzed via standard agarose gel electrophore-sis (1.5%) to confirm presence of an insert. Only confirmed clones were selected for DNA sequencing using standard T7 sequencing primer (T7 RNA polymerase promoter). DNA sequences were screened for identity and genomic origin with the "Basic Local Alignment Search Tool" (BLAST; http://www.ncbi.nlm.nih.gov/ blast/). To determine a potential driving-force for eccD-NA mobilization, all confirmed sequences were analyzed with the repetitive sequence screening tool CEN-SOR [16]; (http://www.girinst.org/censor/index.php).

Results and Discussion
While yeast SIR2 has a stabilizing effect on the organism's rDNA locus, the exact role of SIRT1 on human eccrDNA dynamics is unknown. Interestingly, SIRT1 associates with the human rna45s5 locus which it transcriptionally regulates via the energy-sensing eNoSC protein complex [15,17]. To draw first conclusions about rn5s1, we determined the transcriptional activity of rn5s1 and rna45s5, as well as sirt1, in HEK-293 cells in response to resveratrol (non-significant trends in MCF7 & SH-SY5Y) ( Figure 1A). Consistent with previous work [3,4], we found relevant transcriptional increases of sirt1 itself (1.9-fold) but also rna45s5 (1.8 fold) and rn5s1 (3.5 fold) suggesting parallel regulation and presence of SIRT protein at both rDNA loci.
In gain-of-function mouse models of disease, over-expression of SIRT1 increases homologous recombination of the entire rodent genome [18]. This suggests that DNA stability/mobility depends, in part, on the activation of SIRT1-dependent protein complexes and signaling pathways, particularly in those cells involved in nutrient metabolism and cellular growth. Proper analysis and quantification of eccrDNAs is currently cumbersome -mainly due to their relatively low abundance in human cells and tissues [7]. We therefore sought to develop a sensitive QPCR-based approach with appropriate internal standards, suited for linearized eccrDNA molecules. The sstI family for satellite repeats consists of 2.5 kb repeating units with approximately 400 copies within genomic clusters [19]. The most common alu repeats (1 million copies per haploid genome) are approximately 280 bp long and their average genomic separation of only about 3 kb increases the potential for genomic re-arrangement and e ccD NA formation in mammalian cells. Such non-tandem repeats were previously detected in ecc DNA pools from human cells [20,21]. Whereas alu-eccDNA was reliably detectable across nuclear and cytoplasmic compartments, ss-tI-eccDNA was detectable in most nuclei but rarely in cytoplasm (data not shown). To ensure integrity of the internal standard in response to resveratrol (50 µM), cells were separated into their cytoplasmic and nuclear fractions and processed for isolation of eccDNA and RNA (reversely transcribed into cDNA). The quotient of alu-eccDNA and β-actin expression of the same cells (relative alu-eccDNA index) was independently derived for resveratrol-treated and untreated cells and com-pared. Resveratrol treatment did not significantly affect the concentration of alu-eccDNA in either compartment ( Figure 1B) supporting its suitability as internal standard.
Using our new technique, we tested if the modest resveratrol-dependent transcriptional changes in rn5s1 and rna45s5 were reflected by nuclear and cytoplasmic patterns of the corresponding eccrDNAs (Figure 2A-C). Whereas the average cytoplasmic rn5s1-eccrDNA concentration increased by 2.2-fold (HEK-293), 3.4-fold (MCF7) and 1.5-fold (SH-SY5Y), respectively, nuclear rn5s1-eccDNA was significantly decreased by about 60% (HEK-293), but unaffected in the other two cell lines. Changes in rna45s5-eccrDNA never reached statistical relevance (data not shown).
To confirm the presence of rn5s1-eccrDNA in HEK-293 cells, we analyzed subcellular eccDNA extracts amplified with long-range PCR through Southern-blot analysis ( Figure 2D) using a previously published and labeled rn5s1-eccrDNA probe for detection [7,14]. Initially in this process, all eccDNA molecules were linearized/fragmented with sonication or BamHI restriction (rn5s1-eccrDNA has a unique BamHI restriction site). All samples hybridized in the 2.2 kb range which is consistent with an rn5s1-eccrDNA monomer [12]. Cytoplasmic but not nuclear samples additionally hybridized at approximately 12 kb. Since our protocol includes an initial filtration step, higher molecular weight DNA could have been only extracted in a supercoiled or otherwise condensed state. The signal specificity indicates multicopy rn5s1eccDNA, possibly due to recombination-dependent concatemeric DNA replication as previously described for Drosophila [22]. Both, the resistance to full BamHI restriction and cytoplasm-specific location of higher-order rn5s1-ec-crDNA need to be addressed by future studies.
To further investigate the observed resveratrol-induced cytoplasmic rn5s1-eccrDNA increase in HEK-293 cells, we sought to distinguish between nuclear import and cytoplasmic replication. We thus implemented a 12h treatment regimen involving combinations of resveratrol (50 µM) and WGA (0.1mg/ml) to block nuclear transport, and control lectin ConA (0.1 mg/ml) ( Figure 3A and B). For resveratrol alone and in combination with ConA, we measured respective cytoplasmic rn5s1-eccDNA increases of 180% and 100% (P=0.01) above control levels. WGA alone significantly reduced cytoplasmic rn5s1-eccrDNA to about 50%, whereas WGA plus resveratrol restored levels to 80% baseline. At the same time, resveratrol alone depleted nuclear rn5s1-eccrDNA significantly to about 30% (P=0.01) of the control baseline. Treatment with WGA or WGA plus resveratrol resulted in rescue tendencies toward baseline which were, however, not statistically significant. These results indicate transport of rn5s1-eccrDNA across the nuclear envelope but do not rule out potential compound tributaries through cytoplasmic rn5s1-eccDNA amplification.
Furthermore, we captured 48 additional eccDNA molecules (average: 636 bp) without chromosomal prevalence and with only 1 duplication. While the majority of those eccDNAs were associated with non-repetitive genomic loci, three of them mapped to tandemly arranged gene clusters: rn5S1 (98 repeat average), pcdhα1 (15 repeats) and spanx4 (5 repeats) and another three were mitochondrial ( Figure 4). Out of 45 non-mitochondrial sequences, 30 were intergenic, 12 mapped to intron/ regulatory region boundaries and 3 to coding exon/intron boundaries with further details summarized in Table 1.
Through further analysis with the repetitive DNA sequence mining tool CENSOR [16], we found that 33 of the 45 genomic sequences included 1-4 short repetitive DNA elements (averages 1.7). Of those, 91% were classified as transposable elements encompassing class-I retrotransposons [long interspersed nuclear elements (LINEs; includes l1); short interspersed nuclear elements (SINEs; most abundant class in mammals; includes aluJb, aluJo, aluSc); endogenous retroviruses (ERV1 & 2)] and class-II non-autonomous DNA hAT superfamily transposons ( Table 1). The abundance of these genomic elements is known to facilitate homologous recombination [23] and could therefore contribute to the mobilization of eccDNA. Three of the eccD-NA clones that overlapped with actual coding exons did not include any such repetitive DNA elements and were therefore classified as recently discovered microDNAs [24]. While mechanisms are still at large, our findings are in line with a paradigm based on DNA repetitiveness and mobilization. Our findings confirm the nature of previously described human eccDNA pools and validate our novel experimental approach.
In conclusion, we are introducing a novel and fast PCR-based approach facilitating the isolation, identification and relative quantification of extra-chromosomal circular DNA (eccDNA) obtained from cultured cells. The sensitivity and simplicity of this new methodology permit the study of the subcellular distribution and dynamics of specific eccDNA molecules in response to investigator-selected stimuli (i.e. pharmaceuticals) in a high-throughput fashion. The combination with standard recombinant DNA technology facilitates the identification of novel eccDNA molecules, such as the recently discovered class of coding mi-coDNAs, under desired experimental conditions. This approach therefore provides a new set of tools to be used in the study of the physiology and pathology of eccDNA in general and eccrDNA in particular.  Table 1). The majority of the non-mitochondrial clones (31 with tandem repeats and 3 with unique sequences) contain 1-4 short repetitive DNA elements (with an average of 1.7) that are potentially responsible for their mobilization. The majority (91%) encode transposable elements with class-I retrotransposons (94%; LINE, SINE, erv and alu elements) being dominant over class-II DNA transposons (6%; hAT). The smaller fraction (9%) includes interspersed repeats such as aluS and aluJ.